---
modelId: "tabiclv2"
name: "TabICLv2"
fullName: "TabICLv2 (Inria SODA)"
org: "Inria SODA"
license: "Open Source (BSD-3-Clause)"
elo: "932 (±63)"
rank: 4
browserExecution: "Native WebGPU Browser Runtime (Carla Engine / ONNX Runtime WASM)"
canonicalUrl: "https://carlahq.eu/datasets/models/tabiclv2"
---

# TabICLv2 (Inria SODA) Benchmark Profile

> Inria SODA's state-of-the-art tabular foundation model engineered for instantaneous in-context learning. TabICLv2 runs zero-shot inference without iterative gradient descent and is fully optimized for client-side WebGPU execution in Google Sheets via Carla HQ.

## Technical Specifications
| Dimension | Specification |
| :--- | :--- |
| **Global Leaderboard Rank** | **Rank #4** (Bradley-Terry Elo: 932 (±63)) |
| **Organization** | Inria SODA |
| **Architecture** | In-Context Tabular Transformer with Prefix Attention & KV Caching |
| **Pretraining Prior** | Synthetic Copula-based Quantile Priors & Prior.py DAGs |
| **Context Window** | 1,000+ context rows with zero fine-tuning required |
| **In-Browser Execution** | Native WebGPU Browser Runtime (Carla Engine / ONNX Runtime WASM) |
| **Open-Source License** | Open Source (BSD-3-Clause) |
| **Source Repository** | [TabICLv2 GitHub](https://github.com/soda-inria/tabicl) |
| **Academic Paper** | [Research Paper](https://arxiv.org/abs/2602.11139) |

## Empirical Performance Across Benchmark Datasets
| Dataset | Task Type | Metric | Score | Latency |
| :--- | :--- | :--- | :--- | :--- |
| [Abalone](/datasets/undefined.md) | `regression` | undefined | **2.03** | — |
| [Adult](/datasets/undefined.md) | `binary_classification` | undefined | **0.0793** | — |
| [Airfoil Self Noise](/datasets/undefined.md) | `regression` | undefined | **1.12** | — |
| [Amazon Employee Access](/datasets/undefined.md) | `binary_classification` | undefined | **0.1483** | — |
| [Bank Marketing](/datasets/undefined.md) | `binary_classification` | undefined | **0.1970** | — |
| [Blood Transfusion Service Center](/datasets/undefined.md) | `binary_classification` | undefined | **0.2446** | — |
| [Breast W](/datasets/undefined.md) | `binary_classification` | undefined | **0.0052** | — |
| [Car](/datasets/undefined.md) | `multiclass_classification` | undefined | **0.0231** | — |
| [Compas Two Years](/datasets/undefined.md) | `binary_classification` | undefined | **0.2701** | — |
| [Credit G](/datasets/undefined.md) | `binary_classification` | undefined | **0.2016** | — |
| [Diabetes](/datasets/undefined.md) | `binary_classification` | undefined | **0.1632** | — |
| [Employee Salaries](/datasets/undefined.md) | `regression` | undefined | **8,126** | — |
| [Fitness Club](/datasets/undefined.md) | `binary_classification` | undefined | **0.1795** | — |
| [House Sales](/datasets/undefined.md) | `regression` | undefined | **111,524** | — |
| [Houses](/datasets/undefined.md) | `regression` | undefined | **40,971** | — |
| [Monks Problems 2](/datasets/undefined.md) | `binary_classification` | undefined | **0.0000** | — |
| [Phoneme](/datasets/undefined.md) | `binary_classification` | undefined | **0.0278** | — |
| [Spambase](/datasets/undefined.md) | `binary_classification` | undefined | **0.0075** | — |
| [Telco Customer Churn](/datasets/undefined.md) | `binary_classification` | undefined | **0.1488** | — |
| [Titanic](/datasets/undefined.md) | `binary_classification` | undefined | **0.1229** | — |
| [Vehicle](/datasets/undefined.md) | `multiclass_classification` | undefined | **0.2401** | — |